Application of Classification Models to Pharyngeal High-Resolution Manometry, 2012
Automated Data Analysis of Pharyngeal Pressure, 2011
(with Timothy M. McCulloch, MD; Michelle Ciucci, PhD)
Classification models, including the artificial neural networks (ANNs) multilayer perceptron (MLP) and learning vector quantization (LVQ), as well as support vector machines (SVM), were evaluated for their ability to identify disordered swallowing. Data were collected from 12 control subjects and 13 subjects with swallowing disorders; for this experiment, these subjects swallowed 5-ml water boluses. Following extraction of relevant parameters, a subset of the data was used to train the models, and the remaining swallows were then independently classified by the networks. All methods produced high average classification accuracies, with MLP, SVM, and LVQ achieving accuracies of 96.44%, 91.03%, and 85.39%, respectively. When evaluating the individual contributions of each parameter and groups of parameters to the classification accuracy, parameters pertaining to the upper esophageal sphincter were most valuable.
In addition, we developed an algorithm in MATLAB that can be applied to both normal and disordered swallowing to automatically extract a wide array of measurements from the spatiotemporal plots produced by high-resolution manometry (HRM) of the pharyngeal swallow.
The algorithm was developed from data from 12 normal and 3 disordered subjects swallowing 5-ml water boluses. Automated extraction was compared to manual extraction for a subset of seven normal and the three disordered subjects to evaluate algorithm accuracy. Area and line integrals, pressure wave velocity, and pressure gradients during upper esophageal sphincter opening were also measured. Automated extraction showed strong correlations with manual extraction, producing high correlation coefficients in both normal and disordered subjects for maximum velopharyngeal pressure and maximum tongue base pressure. Timing data were also strongly correlated for all variables, including velopharyngeal pressure duration, tongue base pressure duration, and total swallow duration. Preliminary descriptive data on area and line integrals are presented. Our results indicate that the algorithm can effectively extract data automatically from HRMspatiotemporal plots. The efficiency of the algorithm makes it a valuable tool to supplement clinical and research use of HRM.